Data source: Historical chart pattern similarity via Chart Library API

[grahammccain]

Summary

Chart Library provides historical chart pattern similarity search — given a ticker and date, it finds the 10 most similar historical chart patterns from a database of 24M+ embeddings across 19K symbols and returns what happened next (forward returns at 1/3/5/10 days). This seems like a natural data source for OpenBB's financial analysis platform.

How it would integrate

Chart Library exposes a REST API and an MCP server (pip install chartlibrary-mcp). For OpenBB, the REST API is the most relevant:

import requests

# Find historical patterns similar to NVDA on 2025-01-15
resp = requests.get("https://chartlibrary.io/api/v1/search", params={
    "symbol": "NVDA",
    "date": "2025-01-15",
    "timeframe": "RTH"
}, headers={"X-API-Key": "your-key"})

matches = resp.json()["matches"]
# Each match includes: symbol, date, distance score, 
# forward returns (1d, 3d, 5d, 10d), similarity rank

# Get AI-generated summary of what the pattern implies
summary = requests.get("https://chartlibrary.io/api/v1/summary", params={
    "symbol": "NVDA",
    "date": "2025-01-15"
}, headers={"X-API-Key": "your-key"})

Additional endpoints that could enrich OpenBB data:

• /api/v1/volume-profile/{symbol} — intraday volume vs historical average
• /api/v1/anomaly/{symbol} — pattern anomaly detection (how unusual is today's price action)
• /api/v1/sector-rotation — sector ETF rankings by momentum
• /api/v1/scenario — conditional forward returns ("what if SPY drops 3%?")
• /api/v1/correlation-shift — stocks decorrelating from SPY

Potential integration point

This could work as an OpenBB data provider extension, adding a "pattern similarity" dimension to equity analysis alongside fundamentals and technicals. The API returns structured JSON that maps well to OpenBB's data model.

Details

• Free tier: 200 API calls/day (no credit card)
• Docs: https://chartlibrary.io/developers
• MCP server: pip install chartlibrary-mcp (for agent workflows)
• Coverage: 19K US equities, 10 years of data, 8 timeframes (RTH, premarket, 5min through 5-day)

Happy to help with implementation if there's interest.

[Agostynah]

This looks interesting. How do you account for different market regimes when matching historical patterns? It seems like similarity alone might lead to noisy or misleading analogs.

[dhruvv16-hash]

Hello everyone,

I have completed the implementation of the ChartLibrary pattern similarity search provider integration for the OpenBB Platform.

Details of the Implementation on My Fork:

The complete solution has been pushed to the feature branch on my fork:
https://github.com/dhruvv16-hash/OpenBB/tree/feat/chart-pattern-similarity

The changes include:

1. Standard Models (ChartPatternSimilarity): Added standard query parameters (symbol, date, timeframe, limit) and response structure fields (mapping similarity score, rank, distance, and returns for 1d, 3d, 5d, 10d).
2. Technical Extension Registration: Exposed the endpoint dynamically at obb.technical.pattern_similarity within openbb_technical/technical_router.py.
3. Data Provider Package (openbb-chartlibrary): Created the full provider implementation parsing data requirements and executing POST queries to the chartlibrary.io/api/v1/search/text API.
4. Resiliency/Design Improvements:
   • Mapped the authentication credentials seamlessly using OpenBB settings.
   • Refactored potential fallbacks to ensure valid zero-rank values are correctly parsed.
   • Fully isolated testing routines using mock responses for unit tests (test_chartlibrary_fetchers.py).
5. System Verification: Rebuilt package signatures successfully via poetry run openbb-build and confirmed all 7 test cases pass cleanly.